Accessibility retrofits often arrive in post-production as a deceptively small request: “Can you add captions?” By then, the product may already be back in the factory, the voice actor released, and every shot cut too tightly to describe what is happening on screen. Captions matter, but they solve only one part of the delivery problem.
The European Accessibility Act has applied to consumer e-commerce services since 28 June 2025. The scope is not a blanket claim that every product video has one identical legal obligation. The Directive contains exclusions for pre-recorded time-based media published before that date and for third-party content the operator neither funds, develops nor controls; microenterprises providing services are exempt, while fundamental-alteration and disproportionate-burden provisions also exist. National implementation and the facts of a particular service still matter.
This is therefore a production and procurement framework, not legal advice. Its premise is simple: when video helps a European consumer discover, compare or buy a product, accessibility should be designed into the script, audio, text and product-data handoff—not bolted onto the final MP4.
Five assets that should not be collapsed into “subtitles”
Captions are a synchronized text version of speech and the non-speech audio needed to understand the content. The W3C Web Accessibility Initiative distinguishes them from translated subtitles, which may carry dialogue into another language but do not necessarily include speaker identification or meaningful sound cues.
A basic transcript turns speech and essential non-speech audio into continuous text. A descriptive transcript also includes visual information needed to understand the video: model numbers on screen, the order of assembly steps, a warning light or the state of a product interface.
Audio description makes important visual information available through sound. It can be woven into the main narration, supplied as a separate description track, or delivered in a separate described version.
These are conditional assets. A video whose only sound is background music does not need invented captions; viewers can be told that no meaningful audio is present. A narrated demonstration that already verbalizes every important visual action may not need a redundant description track. The buying decision should follow the information, not a fixed file checklist.
ONCE’s four-layer accessible product-video delivery
1. Meaning: no purchase-critical fact should live in only one sense
Start by asking what decision a viewer loses if they cannot hear the soundtrack, see the image clearly, or read text embedded in the frame.
If a blinking green light confirms pairing, “it will look like this” is not enough. The narration can say, “Press and hold the side button until the green indicator blinks.” If compatible models appear only on packaging, those models also belong in narration, a descriptive transcript or accessible page text. A warning conveyed only by red colour, or an alert conveyed only by sound, needs another perceptible form.
This audit belongs in the script and storyboard. It determines the rest of the deliverables.
2. Hearing access: supply editable caption assets, not only burned-in text
Open captions are useful in silent social feeds, but an e-commerce page or accessible player needs a file that can be toggled, translated and processed by assistive technology. For each approved language, the practical package should include:
- a human-reviewed WebVTT or SRT file;
- same-language captions with relevant speaker and sound cues;
- a social-safe open-caption version where required;
- a text transcript derived from the reviewed caption source.
Automatic captions are a draft, not an acceptance criterion. YouTube’s own guidance says machine-generated captions can misrepresent speech because of pronunciation, accents, dialects and background noise, and should always be reviewed. Product QA should check brand names, model numbers, units, negations, technical terms, overlapping speakers, meaningful sounds, line breaks and synchronization—not merely spelling.
3. Vision access: describe important actions in the main script first
For a new how-to, installation guide or product demonstration, the most economical route is often integrated description. W3C’s description guidance recommends putting relevant visual information into the main speakers’ scripts when the format permits.
“Attach this here” becomes “Attach the small metal ring to the green, larger end.” “The interface changes” becomes “A blue connection icon appears at the top of the display.” This helps blind and low-vision viewers, but it also reduces ambiguity in audio-first playback, distant viewing and multilingual dubbing.
When story pacing, continuous dialogue or dense visual action makes integrated description impractical, choose a timed description file, separate audio track, described version or descriptive transcript. The later that decision is made, the more likely the edit will need new recording, longer shots and a rebuilt mix.
4. Commerce data: keep essential product facts accessible outside the video
Price, compatibility, dimensions, cautions, warranty and qualifying conditions should never exist only inside a video. The film can demonstrate and persuade; accessible page text and structured product data must remain readable, updatable and auditable.
This complements the ONCE product truth pack: the truth pack keeps claims traceable, while accessible delivery makes the same facts obtainable through audio, text, visual description and page fields. Both need language, market, version, effective date, owner and retirement status.
A four-question decision path before the shoot
- Does the soundtrack carry unique meaning? If yes, commission accurate synchronized captions that include necessary non-speech audio. If not, state that the soundtrack is only background music.
- Does the image carry purchase information the narration omits? Rewrite the main narration first; if that cannot work, specify a description track, described version or descriptive transcript.
- Will the information change by market, SKU or date? Do not lock it only into the picture. Maintain accessible page text and versioned product fields.
- How many languages, channels and players will carry the asset? Verify what each destination supports before approving masters, caption files and fallback versions. One MP4 is not a delivery strategy.
A useful acceptance package can therefore include a clean master, social-safe version, WebVTT/SRT for every language, basic or descriptive transcript, integrated-description sign-off or separate description script and track, accessible product-fact copy, and an inventory recording language, market, channel, version, date, owner and status.
Compare accessible combinations, not caption minutes
Instead of relying on a made-up industry average, ONCE recommends an internal variable:
Cost per accepted accessible distribution combination = (script restructuring + captions/transcripts + descriptive audio/text + multilingual QA + player/page adaptation + version maintenance) ÷ accepted language × channel × media-form combinations.
This reveals why a slightly longer pre-production step can lower the total. Integrated narration may remove a separate description recording and several localization fixes. A cheaper package containing only burned-in captions may have to be rebuilt when the video reaches a product page, another language or a different player.
Captions are not the finish line. The valuable deliverable is a maintainable system that keeps product meaning consistent across sound, text, visual description and commerce data. For brands sourcing production through China, that is also the moment to resolve model names, interface language, units and assembly actions while the product and the people who understand it are still in the room.
